Related papers: A High-Throughput Hardware Accelerator for Lempel-…
The pressing need for efficient compression schemes for XML documents has recently been focused on stack computation, and in particular calls for a formulation of information-lossless stack or pushdown compressors that allows a formal…
We propose algorithms computing the semi-greedy Lempel-Ziv 78 (LZ78), the Lempel-Ziv Double (LZD), and the Lempel-Ziv-Miller-Wegman (LZMW) factorizations in linear time for integer alphabets. For LZD and LZMW, we additionally propose data…
Power density constraints are limiting the performance improvements of modern CPUs. To address this we have seen the introduction of lower-power, multi-core processors such as GPGPU, ARM and Intel MIC. To stay within the power density…
Compression refers to encoding data using bits, so that the representation uses as few bits as possible. Compression could be lossless: i.e. encoded data can be recovered exactly from its representation) or lossy where the data is…
Compression techniques that support fast random access are a core component of any information system. Current state-of-the-art methods group documents into fixed-sized blocks and compress each block with a general-purpose adaptive…
As a core component in modern data centers, key-value cache provides high-throughput and low-latency services for high-speed data processing. The effectiveness of a key-value cache relies on its ability of accommodating the needed data.…
In the era of diminishing returns from Moores Law, heterogeneous computing systems have emerged as a vital approach to enhance computational efficiency. This paper introduces a novel MLIR-based dialect, named hyper, designed to optimize…
Hardware accelerators, in particular accelerators for tensor processing, have many potential application domains. However, they currently lack the software infrastructure to support the majority of domains outside of deep learning.…
Specialized hardware like application-specific integrated circuits (ASICs) remains the primary accelerator type for cryptographic kernels based on large integer arithmetic. Prior work has shown that commodity and server-class GPUs can…
Large-scale models rely heavily on 3D parallelism for distributed training, which utilizes tensor parallelism (TP) as the intra-operator parallelism to partition model states across GPUs. However, TP introduces significant communication…
Relative Lempel-Ziv (RLZ) parsing is a dictionary compression method in which a string $S$ is compressed relative to a second string $R$ (called the reference) by parsing $S$ into a sequence of substrings that occur in $R$. RLZ is…
Sequence model based NLP applications can be large. Yet, many applications that benefit from them run on small devices with very limited compute and storage capabilities, while still having run-time constraints. As a result, there is a need…
HipMCL is a high-performance distributed memory implementation of the popular Markov Cluster Algorithm (MCL) and can cluster large-scale networks within hours using a few thousand CPU-equipped nodes. It relies on sparse matrix computations…
With the wide adoption of language models for IR -- and specifically RAG systems -- the latency of the underlying LLM becomes a crucial bottleneck, since the long contexts of retrieved passages lead large prompts and therefore, compute…
The auto-regressive decoding of Large Language Models (LLMs) results in significant overheads in their hardware performance. While recent research has investigated various speculative decoding techniques for multi-token generation, these…
Lempel-Ziv is an easy-to-compute member of a wide family of so-called macro schemes; it restricts pointers to go in one direction only. Optimal bidirectional macro schemes are NP-complete to find, but they may provide much better…
Relative Lempel-Ziv (RLZ) is a popular algorithm for compressing databases of genomes from individuals of the same species when fast random access is desired. With Kuruppu et al.'s (SPIRE 2010) original implementation, a reference genome is…
Although it has been surpassed by many subsequent coding standards, JPEG occupies a large share of the storage load of the current data hosting service. To reduce the storage costs, DropBox proposed a lossless secondary compression…
Point-based 3D point cloud models employ computation and memory intensive mapping functions alongside NN layers for classification/segmentation, and are executed on server-grade GPUs. The sparse, and unstructured nature of 3D point cloud…
Transformers have revolutionized AI in natural language processing and computer vision, but their large computation and memory demands pose major challenges for hardware acceleration. In practice, end-to-end throughput is often limited by…